This paper considers the problem of cross-sectional aggregation when the underlying micro behavioural relations are characterised by general non-linear specifications. It focuses on forecasting the aggregates, and shows how an optimal aggregate model can be derived by minimising the mean squared prediction errors conditional on the aggregate information. It also derives model selection criteria for distinguishing between aggregate and disaggregate models when the primary object of the analysis is forecasting the aggregates, and establishes the consistency of the model selection criteria in large samples. In the case of standard non-linear micro relations with additive specifications, boot-strap techniques are considered to correct for small sample bias of the proposed model selection criteria. The paper also contains an empirical application where log-linear production functions are estimated for the UK economy disaggregated by eight industrial sectors and at the aggregate level for 1954-1995.
